VIDEO MODEL by Decart 14B

Lucy

Decart's fast image-to-video model — animate any still image with physics-aware motion, cinematic quality, and lightning-fast generation. Lucy transforms a single frame and a text prompt into a 10-second 720p video clip with consistent motion and no visual artifacts.

Resolution

720p

Aspect ratios

16:9, 9:16

Base credits

240

Input

Start frame

## Fast image-to-video animation Lucy is Decart's 14-billion parameter image-to-video model, built on a diffusion-transformer architecture that delivers cinematic motion from a single still image. You provide a start frame — a photo, an illustration, a rendered scene — write a prompt describing the motion, and Lucy generates a smooth 720p clip in seconds. What sets Lucy apart is its physics-aware generation. The model learns the structure of the world implicitly — understanding how fabric drapes, how liquids move, how surfaces react to light — without relying on depth maps, green screens, or 3D meshes. The result is motion that looks natural rather than interpolated: subjects move as they would in the real world, and the visual quality holds consistently frame to frame without flickering or drift. Lucy is the fastest option on ImagineArt for creators who need high-quality animated clips from existing imagery — product shots, character art, concept illustrations, or photos — without the overhead of higher-credit models. ## Capabilities Upload any still image as the starting frame — photo, illustration, or rendered scene — and Lucy animates it with natural, physics-consistent motion. The model understands world structure implicitly: fabric moves like fabric, liquids flow realistically, and surfaces respond to light correctly — no 3D rigs or depth data required. Optimized for low-latency inference — Lucy generates clips significantly faster than comparable quality models, making iteration quick and affordable. Describe the motion, camera angle, atmosphere, and style in natural language. Lucy follows complex multi-part instructions and applies them to the starting frame. Frame-to-frame consistency is maintained throughout the clip — eliminating the flickering, morphing, and temporal artifacts common in lower-quality video models. Accepts JPG, JPEG, PNG, WebP, GIF, and AVIF input files — compatible with the full range of image formats used in creative and production workflows. ## Specifications | Feature | Details | | -------------------- | ------------------------------- | | **Developer** | Decart | | **Model size** | 14B parameters | | **Resolution** | 720p | | **Aspect ratios** | 16:9, 9:16 | | **Input** | Start frame (image-to-video) | | **Accepted formats** | JPG, JPEG, PNG, WebP, GIF, AVIF | | **Output format** | MP4 (H.264) | | **Base credits** | 240 | | **Audio** | No native audio generation | ## How to use Log into ImagineArt and go to the **AI Video Generator**. Choose **Lucy** from the model dropdown. Upload the image you want to animate. Lucy uses this as the first frame of the generated clip. Supported formats: JPG, JPEG, PNG, WebP, GIF, AVIF. Describe what you want to happen in the clip — the motion, camera movement, atmosphere, and any stylistic direction. Be specific about how things should move. Choose **16:9** for landscape/widescreen or **9:16** for vertical/portrait content. Click **Generate**. Lucy animates your start frame and delivers a 720p MP4 clip. ## Prompting tips * **Describe motion explicitly** — Lucy needs to know what moves and how. "The camera slowly pushes in," "the subject turns their head to look left," "leaves drift downward" are all more actionable than vague scene descriptions. * **Reference the starting image indirectly** — Lucy already knows what the scene looks like from your start frame. Focus your prompt on motion, camera behavior, and atmosphere rather than restating visual elements. * **Use physics language** — Phrases like "fabric ripples in the breeze," "water surface shimmers," or "steam rises from the cup" take advantage of Lucy's physics-aware generation to produce natural results. * **Aspect ratio determines framing** — Choose 16:9 for landscape subjects (scenes, cityscapes, wide shots) and 9:16 for portrait subjects (people, vertical compositions, social media delivery). * **Keep motion achievable in 10 seconds** — A single, clear action or camera move tends to produce better results than a complex sequence. Save multi-shot narratives for models like Seedance or Wan. ### Example prompts > The subject slowly turns toward the camera, hair catching a gentle breeze. Soft afternoon light. Natural movement, cinematic. 16:9. > A steaming mug of coffee sits on a wooden desk. Gentle wisps of steam rise and curl. Shallow depth of field, warm tones, static camera. 16:9. > A city street at night. Rain begins to fall softly, droplets catching the neon reflections on wet pavement. The camera holds still. 16:9. > A fashion model stands against a white backdrop. Fabric of the dress moves gently as if caught in a slow breeze. 9:16, vertical format. ## Compare models | Model | Input | Resolution | Speed | Best for | | ----------------------------------------------- | ------------- | ----------- | -------- | --------------------------------------- | | **Lucy** | Start frame | 720p | Fast | Quick image animation, social content | | [Luma Ray 2](/ai-models/video/luma-ray-2) | Text or image | Up to 1080p | Moderate | Photorealistic textures, natural motion | | [Kling 2.6 Pro](/ai-models/video/kling-2-6-pro) | Text or image | 1080p | Moderate | Cinematic quality, audio sync | | [Seedance 2](/ai-models/video/seedance-2) | Text or image | 1080p | Moderate | High-fidelity multi-modal generation | | [Runway 4.5](/ai-models/video/runway-4-5) | Text or image | 720p | Moderate | Camera-precise cinematic output | Lucy is the best choice when you already have a strong still image and want to animate it quickly without high credit cost. For productions that need higher resolution, audio, or complex multi-shot control, consider stepping up to Kling, Seedance, or a Wan model.